Installed HA on a Pi4 routerboard a while back. It only has the Pi4 Compute Module and 2 NICs (no USB, HDMI, etc). Access is only possible via network, or the filesystem can be accessed if the the microSD card is removed and read on another system.
It was running fine, but after attempting upgrade it went offline and will not respond to http/https anymore. I can see it is requesting a DHCP address from the router and it will respond to pings. I tried to SSH into it but also failed (I think I have to enable this from the web interface first).
Can anyone recommend any tricks to gain access to it? Otherwise if I remove the SD card is it possible to backup the config from this to restore to another HA once I rebuild it? Are there any suggested things to check on the filesystem of the SD card that might help to get the existing HA accessible again?
Any recommendations are appreciated.
I have noticed that tailscale is still running and active on it, though I can’t access via the web interface on this either. While this doesn’t help me access, it does prove that the underlying system is still running, it’s just the web interface that is broken.
Hoping for any tips from the community.
Sadly I never got any help on this one, so I had to turn to chatGPT.
Here is the result:
I removed the SD card and tested it from a laptop. Windows only saw 1 partition (hassos-overlay) and there was nothing useful in this.
I then booted the laptop from an Ubuntu v24 live USB disk which was able to read the partitions on the SD card a bit better. However, the main partition I wanted was the hassos-data one and this was not visible/mounted.
Running lsblk -f I saw there were 8 partitions there and mmcblk0p8 was the one I needed. I created a dir and mounted into it:
sudo mkdir /mnt/hassdata
sudo mount /dev/mmcblk0p8 /mnt/hassdata
however, this didn’t work .
I also tried it read-only as
sudo mount -o ro /dev/mmcblk0p8 /mnt/hassdata
but also could get no luck with this. Finally I tried
sudo mount -t ext4 -o ro,noload /dev/mmcblk0p8 /mnt/hassdata
and this worked. I was then able to copy the supervisor folder elsewhere so now I can pick through the files and see if I can locate a usable backup. I have built a new HA on a VM in a Proxmox host and I will see if I can find a backup to restore my setup onto this new build.
To close this: The solution was to remove the SD card and read it on a Linux system. The partition containing the backups was corrupted, but it was able to be mounted as read only. As above, I created a folder for the partition to mount to, then did
sudo mount /dev/mmcblk0p8 /mnt/hassdata
sudo mount -t ext4 -o ro,noload /dev/mmcblk0p8 /mnt/hassdata
Then I copied over the entire directory of /supervisor/backups. There were around 60 backups in there. All of them were tar files around 60-80Mb. I moved them to a USB disk, then tested restoring into the HA interface. All my first attempts would fail with the restore seemingly taking forever (where it had really crashed or failed).
Then I tested opening each of the tar files to check if the archives themselves were OK. I was on Windows, so I used 7zip for this. Almost all of the archives failed to open with general errors or CRC errors, so obvioulsy corrupted. However, I was lucky enough to find 1 older backup after about 20 failures that would open in 7zip. Once I had established the good backup, I was able to restore it into HA and it works.